Quantization made by Richard Erkhov.
neuronovo-9B-v0.3 - GGUF
- Model creator: https://huggingface.co/Neuronovo/
- Original model: https://huggingface.co/Neuronovo/neuronovo-9B-v0.3/
Name | Quant method | Size |
---|---|---|
neuronovo-9B-v0.3.Q2_K.gguf | Q2_K | 3.13GB |
neuronovo-9B-v0.3.IQ3_XS.gguf | IQ3_XS | 3.48GB |
neuronovo-9B-v0.3.IQ3_S.gguf | IQ3_S | 3.67GB |
neuronovo-9B-v0.3.Q3_K_S.gguf | Q3_K_S | 3.65GB |
neuronovo-9B-v0.3.IQ3_M.gguf | IQ3_M | 3.79GB |
neuronovo-9B-v0.3.Q3_K.gguf | Q3_K | 4.05GB |
neuronovo-9B-v0.3.Q3_K_M.gguf | Q3_K_M | 4.05GB |
neuronovo-9B-v0.3.Q3_K_L.gguf | Q3_K_L | 4.41GB |
neuronovo-9B-v0.3.IQ4_XS.gguf | IQ4_XS | 4.55GB |
neuronovo-9B-v0.3.Q4_0.gguf | Q4_0 | 4.74GB |
neuronovo-9B-v0.3.IQ4_NL.gguf | IQ4_NL | 4.79GB |
neuronovo-9B-v0.3.Q4_K_S.gguf | Q4_K_S | 4.78GB |
neuronovo-9B-v0.3.Q4_K.gguf | Q4_K | 5.04GB |
neuronovo-9B-v0.3.Q4_K_M.gguf | Q4_K_M | 5.04GB |
neuronovo-9B-v0.3.Q4_1.gguf | Q4_1 | 5.26GB |
neuronovo-9B-v0.3.Q5_0.gguf | Q5_0 | 5.77GB |
neuronovo-9B-v0.3.Q5_K_S.gguf | Q5_K_S | 5.77GB |
neuronovo-9B-v0.3.Q5_K.gguf | Q5_K | 5.93GB |
neuronovo-9B-v0.3.Q5_K_M.gguf | Q5_K_M | 5.93GB |
neuronovo-9B-v0.3.Q5_1.gguf | Q5_1 | 6.29GB |
neuronovo-9B-v0.3.Q6_K.gguf | Q6_K | 6.87GB |
neuronovo-9B-v0.3.Q8_0.gguf | Q8_0 | 8.89GB |
Original model description:
license: apache-2.0 datasets: - Intel/orca_dpo_pairs - mlabonne/chatml_dpo_pairs language: - en library_name: transformers
More information about previous Neuronovo/neuronovo-9B-v0.2 version available here: ๐Don't stop DPOptimizing!
Author: Jan Kocoล ๐LinkedIn ๐Google Scholar ๐ResearchGate
Changes concerning Neuronovo/neuronovo-9B-v0.2:
Training Dataset: In addition to the Intel/orca_dpo_pairs dataset, this version incorporates a mlabonne/chatml_dpo_pairs. The combined datasets enhance the model's capabilities in dialogues and interactive scenarios, further specializing it in natural language understanding and response generation.
Tokenizer and Formatting: The tokenizer now originates directly from the Neuronovo/neuronovo-9B-v0.2 model.
Training Configuration: The training approach has shifted from using
max_steps=200
tonum_train_epochs=1
. This represents a change in the training strategy, focusing on epoch-based training rather than a fixed number of steps.Learning Rate: The learning rate has been reduced to a smaller value of
5e-6
. This finer learning rate allows for more precise adjustments during the training process, potentially leading to better model performance.